prefixspan python

from：https://github.com/chuanconggao/PrefixSpan-py

 

API Usage

Alternatively, you can use the algorithms via API.

from prefixspan import PrefixSpan

db = [
    [0, 1, 2, 3, 4],
    [1, 1, 1, 3, 4],
    [2, 1, 2, 2, 0],
    [1, 1, 1, 2, 2],
]

ps = PrefixSpan(db)

For details of each parameter, please refer to the PrefixSpan class in prefixspan/api.py.

设置长度限制：

ps = PrefixSpan(db)
ps.minlen = 3
ps.maxlen = 5
print("?"*66)
------------------

print(ps.frequent(2))
# [(2, [0]),
#  (4, [1]),
#  (3, [1, 2]),
#  (2, [1, 2, 2]),
#  (2, [1, 3]),
#  (2, [1, 3, 4]),
#  (2, [1, 4]),
#  (2, [1, 1]),
#  (2, [1, 1, 1]),
#  (3, [2]),
#  (2, [2, 2]),
#  (2, [3]),
#  (2, [3, 4]),
#  (2, [4])]

print(ps.topk(5))
# [(4, [1]),
#  (3, [2]),
#  (3, [1, 2]),
#  (2, [1, 3]),
#  (2, [1, 3, 4])]


print(ps.frequent(2, closed=True))

print(ps.topk(5, closed=True))


print(ps.frequent(2, generator=True))

print(ps.topk(5, generator=True))

Closed Patterns and Generator Patterns

一个 频繁的顺序模式 是一种出现在序列数据库的至少“minsup”序列中的模式，其中 最小支持度 是用户设置的参数。

一个 频繁闭合序列模式 是一种频繁的顺序模式，使得它不包括在具有完全相同支持的另一顺序模式中。

算法如 的PrefixSpan 找到频繁的顺序模式。算法如 BIDE+找到频繁的闭合序列模式。 BIDE +通常比PrefixSpan快得多，因为它使用修剪技术来避免生成所有顺序模式。此外，闭合模式集通常比连续模式集小得多，因此BIDE +也更具存储效率。

另一个重要的事情是，闭合序列模式是所有序列模式的紧凑和无损表示。这意味着闭合序列模式的集合通常要小得多，但它是无损的，这意味着它允许恢复整个连续模式集（没有信息丢失），这非常方便。

我可以举个简单的例子。

让我们考虑4个序列：

a  b  c  d  e
a  b  d
b  e  a  
b  c  d  e

让我们说minsup = 2。

b c 是一种频繁的序列模式，因为它出现在两个序列中（它支持2）。 b c 不是一个封闭的顺序模式，因为它包含在一个更大的顺序模式中 b c d 得到同样的支持。

b c d 它也是一个支持2.它也不是一个封闭的顺序模式，因为它包含在一个更大的顺序模式中 b c d e 得到同样的支持。 b c d e 是一个封闭的顺序模式，因为它没有包含在具有相同支持的任何其他顺序模式中。

The closed patterns are much more compact due to the smaller number.

• A pattern is closed if there is no super-pattern with the same frequency.

prefixspan-cli frequent 2 --closed test.dat

0 : 2
1 : 4
1 2 : 3
1 2 2 : 2
1 3 4 : 2
1 1 1 : 2

The generator patterns are even more compact due to both the smaller number and the shorter lengths.

• A pattern is generator if there is no sub-pattern with the same frequency.

• Due to the high compactness, generator patterns are useful as features for classification, etc.

prefixspan-cli frequent 2 --generator test.dat

0 : 2
1 1 : 2
2 : 3
2 2 : 2
3 : 2
4 : 2

There are patterns that are both closed and generator.

prefixspan-cli frequent 2 --closed --generator test.dat

0 : 2

备注：模式挖掘有很多算法。

SPMF offers implementations of the following data mining algorithms.

Sequential Pattern Mining

These algorithms discover sequential patterns in a set of sequences. For a good overview of sequential pattern mining algorithms, please read this survey paper.

• algorithms for mining sequential patterns in a sequence database
  • the CM-SPADE algorithm (Fournier-Viger et al, 2014, powerpoint)
  • the CM-SPAM algorithm (Fournier-Viger et al, 2014, powerpoint)
  • the FAST algorithm (Salvemini et al, 2011) 
  • the GSP algorithm (Srikant et al., 1996)
  • the LAPIN (aka LAPIN-SPAM) algorithm (Yang et al., 2005)
  • the PrefixSpan algorithm (Pei et al., 2004)
  • the SPADE algorithm (Zaki et al., 2001)
  • the SPAM algorithm (Ayres et al., 2002)
• algorithms for mining closed sequential patterns in a sequence database
  • the ClaSP algorithm (Gomariz et al., 2013)
  • the CM-ClaSP algorithm (Fournier-Viger et al, 2014, powerpoint)
  • the CloFAST algorithm (Fumarola et al, 2016) 
  • the CloSpan algorithm (Yan et al., 2003)
  • the BIDE+ algorithm(Wang et al., 2007)
• algorithms for mining maximal sequential patterns in a sequence database
  • the VMSP algorithm (Fournier-Viger et al, 2014, powerpoint)
  • the MaxSP algorithm (Fournier-Viger et al., 2013, powerpoint).
• algorithms for mining the top-k sequential patterns in a sequence database
  • the TKS algorithm (Fournier-Viger et al., 2013, powerpoint).
  • the TSP algorithm (Tzvetkoz et al., 2003).
  • the Skopus algorithm for mining the top-k sequential patterns using leverage and significance (Petijean et al., 2016)
• algorithms for mining sequential generator patterns in a sequence database
  • the VGEN algorithm (Fournier-Viger et al, 2014)
  • the FEAT algorithm (Gao et al., 2008).
  • the FSGP algorithm (Yi et al., 2011).
• algorithms for mining compressing sequential patterns
  • the GoKrimp and SeqKrimp algorithms (Lam et al., 2012; Lam et al., 2014)
• algorithms for mining multidimensional sequential patterns in a multidimensional sequence database
  • the SeqDIM algorithm for mining frequent multidimensional sequential patterns in a multi-dimensional sequence database (Pinto et al., 2001)
  • the Songram et al. algorithm for mining frequent closed multidimensional sequential patterns in a multi-dimensional sequence database (Songram et al. 2006)
• the Fournier-Viger et al. algorithm, a sequential pattern mining algorithm that combines several features from well-known sequential pattern mining algorithms and also proposes some original features (Fournier-Viger et al., 2008):
  • mining sequences with minimum support by database-projection (based on PrefixSpan, Pei et al., 2004)
  • mining sequences with min/max time interval between events and min/max time length of a sequence (based on Hirate-Yamana, 2006)
  • mining closed sequences (based on the BIDE+ algorithm by Wang et al. 2007)
  • mining multi-dimensional sequences (based on Pinto et al. 2001)
  • mining closed multi-dimensional sequences (based on Songram et al. 2006 and Pasquier et al., 1999)
  • mining sequences with items having integer values and performing automatic clustering of these values (original extension described in Fournier-Viger et al., 2008)
• algorithm for mining high-utility sequential patterns in a sequence database 
  • the USPAN algorithm (Yin et al. 2012)
• algorithm for mining high-utility probability sequential patterns in a sequence database 
  • the PHUSPM algorithm (Zhang et al. 2018) 
  • the UHUSPM algorithm (Zhang et al. 2018) 
• algorithm for progressive sequential pattern mining with convergence guarantees
  • the ProSecCo algorithm (Servan-Schreiber et al. 2018) 
• the Occur algorithm for finding all occurrences of some sequential patterns in sequences by post-processing.

Sequential Rule Mining

These algorithms discover sequential rules in a set of sequences.

• algorithms for mining sequential rules in a sequence database
  • the ERMiner algorithm (Fournier-Viger et al., 2014)
  • the RuleGrowth algorithm (Fournier-Viger et al., 2011, Fournier-Viger et al., 2015, powerpoint, video)
  • the CMRules algorithm (Fournier-Viger et al., 2010, powerpoint)
  • the CMDeo algorithm (Fournier-Viger et al., 2010)
  • the RuleGen algorithm (Zaki et al, 2001)
• algorithms for mining sequential rules in a sequence database with the window size constraint
  • the TRuleGrowth algorithm (Fournier-Viger, 2012a, Fournier-Viger et al., 2015)
• algorithms for mining top-k sequential rules in a sequence database
  • the TopSeqRules algorithm for mining the top-k sequential rules (Fournier-Viger et al., 2011, powerpoint)
  • the TopSeqClassRules algorithm for mining the top-k class sequential rules (a variation of Fournier-Viger et al., 2011)
  • the TNS algorithm for mining the top-k non-redundant sequential rules (Fournier-Viger 2013)
• algorithm for mining high-utility sequential rules in a sequence database 
  • the HUSRM algorithm (Zida et al., 2015)

Sequence Prediction

These algorithms predict the next symbol(s) of a sequence based on a set of training sequences

• algorithms for predicting the next symbol of a sequence based on a set of training sequences
  • the Compact Prediction Tree+ (CPT+) algorithm (Gueniche et al., 2015, powerpoint)
  • the Compact Prediction Tree (CPT) algorithm (Gueniche et al., 2013)
  • the First order Markov Chains (PPM - order 1) (Clearly et al, 1984)
  • the Dependency Graph (DG) (Padmanabhan, 1996)
  • the All-k-Order Markov Chains (AKOM) (Pitkow, 1999)
  • the TDAG (Laird & Saul, 1994)
  • the LZ78 (Ziv, 1978)

Itemset Mining

These algorithms discover interesting itemsets (sets of values) that appear in a transaction database (database records containing symbolic data). For a good overview of itemset mining, please read this survey paper.

• algorithms for discovering frequent itemsets in a transaction database.
  • the Apriori algorithm (Agrawal & Srikant, 1994)
  • the AprioriTID algorithm (Agrawal & Srikant, 1994)
  • the FP-Growth algorithm (Han et al., 2004)
  • the Eclat algorithm (Zaki, 2000)
  • the dEclat algorithm (Zaki and Gouda, 2001, 2003)
  • the Relim algorithm (Borgelt, 2005)
  • the H-Mine algorithm (Pei et al., 2007)
  • the LCMFreq algorithm (Uno et al., 2004)
  • the PrePost and PrePost+ algorithms (Deng et al., 2012, Deng et Lv, 2015)
  • the FIN algorithm (Deng et al., 2014)
  • the DFIN algorithm (Deng et al., 2016)
  • the NegFIN algoritm (Aryabarzan et al., 2018) 
• algorithms for discovering frequent closed itemsets in a transaction database.
  • the FPClose algorithm (Grahne and Zhu, 2005)
  • the Charm algorithm (Zaki and Hsiao, 2002)
  • the dCharm algorithm (Zaki and Gouda, 2001)
  • the DCI_Closed algorithm (Lucchese et al, 2004)
  • the LCM algorithm (Uno et al., 2004)
  • the AprioriClose aka Close algorithm (Pasquier et al., 1999)
  • the AprioriTID Close algorithm (Pasquier et al., 1999, Agrawal & Srikant, 1994)
• algorithms for recovering all frequent itemsets from frequent closed itemsets:
  • the LevelWise algorithm (Pasquier et al., 1999) 
  • the DFI-Growth algorithm (___ et al., 2018) 
• algorithms for discovering frequent maximal itemsets in a transaction database.
  • the FPMax algorithm (Grahne and Zhu, 2003)
  • the Charm-MFI algorithm for discovering frequent closed itemsets and maximal frequent itemsets by post-processing in a transaction database (Szathmary et al. 2006)
• algorithms for mining frequent itemsets with multiple minimum supports
  • the MSApriori algorithm (Liu et al, 1999)
  • the CFPGrowth++ algorithm (Uday & Reddy, 2011, Hu & Chen, 2006)
• algorithms for mining generator itemsets in a transaction database
  • the DefMe algorithm for mining frequent generator itemsets in a transaction database (Soulet & Rioult, 2014)
  • the Pascal algorithm for mining frequent itemsets, and identifying at the same time which one are generators (Bastide et al., 2002)
  • the Zart algorithm for discovering frequent closed itemsets and their generators in a transaction database (Szathmary et al. 2007)
• algorithms for mining rare itemsets and/or correlated itemsets in a transaction database
  • the AprioriInverse algorithm for mining perfectly rare itemsets (Koh & Roundtree, 2005)
  • the AprioriRare algorithm for mining minimal rare itemsets and frequent itemsets (Szathmary et al. 2007b)
  • the CORI algorithm for mining minimal rare correlated itemsets using the support and bond measures (Bouasker et al. 2015)
  • the RP-Growth algorithm for mining rare itemsets (Tsang et al., 2011) 
• algorithms for performing targeted and dynamic queries about association rules and frequent itemsets.
  • the Itemset-Tree, a data structure that can be updated incrementally, and algorithms for querying it. (Kubat et al, 2003)
  • the Memory-Efficient Itemset-Tree, a data structure that can be updated incrementally, and algorithms for querying it. (Fournier-Viger, 2013, powerpoint)
• algorithms to discover frequent itemsets in a stream
  • the estDec algorithm for mining recent frequent itemsets in a data stream (Chang & Lee, 2003)
  • the estDec+ algorithm for mining recent frequent itemsets in a data stream (Shin et al., 2014)
  • the CloStream algorithm for mining frequent closed itemsets in a data stream (Yen et al, 2009)
• the U-Apriori algorithm for mining frequent itemsets in uncertain data (Chui et al, 2007)
• the VME algorithm for mining erasable itemsets (Deng & Xu, 2010)
• algorithms to discover fuzzy frequent itemsets in a quantitative transaction database
  • the FFI-Miner algorithm for mining fuzzy itemsets (Lin et al., 2015)
  • the MFFI-Miner algorithm for mining multiple fuzzy itemsets (Lin et al., 2016)

Periodic Pattern Mining

These algorithms discover patterns that periodically appear in a sequence of complex events (also called a transaction database)

• the PFPM algorithm (Fournier-Viger et al, 2016a, powerpoint, video  ) for mining frequent periodic patterns in a sequence of transactions (a transaction database))
• the PHM algorithm (Fournier-Viger et al, 2016b, powerpoint) for mining periodic high-utility patterns (periodic patterns that yield a high profit) in a sequence of transactions (a transaction database) containing utility information 

Episode Mining

These algorithms discover episodes that appear in a single sequence of complex events.

• the TUP algorithm (Rathore et al., 2016) for mining the top-k high utility episodes in a sequence of complex events (a transaction database) with utility information 
• the US-SPAN algorithm (Wu et al., 2013 ) for mining high utility episodes in a sequence of complex events (a transaction database) with utility information 

High-Utility Pattern Mining

These algorithms discover patterns having a high utility (importance) in different kinds of data. For a good overview of high utility itemset mining, you may read this survey paper, and the high utility-pattern mining book.

• algorithms for mining high-utility itemsets in a transaction database having profit information
  • the EFIM algorithm (Zida et al. 2016, Zida et al., 2015, powerpoint)
  • the FHM algorithm (Fournier-Viger et al., 2014, powerpoint)
  • the HUI-Miner algorithm (Liu & Qu, 2012)
  • the HUP-Miner algorithm (Krishnamoorthy, 2014)
  • the mHUIMiner algorithm (Peng et al., 2017)
  • the HMiner algorithm (Krishnamoorty, 2017)
  • the ULB-Miner algorithm (Duong et al, 2018) 
  • the UFH algorithm (Dawar et al, 2017) 
  • the IHUP algorithm (Ahmed et al., 2009)
  • the Two-Phase algorithm (Liu et al., 2005)
  • the UP-Growth algorithm (Tseng et al., 2011)
  • the UP-Growth+ algorithm (Tseng et al., 2013)
  • the UP-Hist algorithm (Dawar et al., 2015) 
  • the d2HUP algorithm (Liu et al, 2012)
• algorithm for efficiently mining high-utility itemsets with length constraints in a transaction database
  • the FHM+ algorithm (Fournier-Viger et al, 2016, powerpoint)
• algorithm for mining correlated high-utility itemsets in a transaction database
  • the FCHM_bond algorithm, to use the bond measure (Fournier-Viger et al, 2016, powerpoint, Fournier-Viger 2018 et al., to appear, video )
  • the FCHM_allconfidence algorithm, to use the all-confidence measure (Fournier-Viger et al, 2016, powerpoint, Fournier-Viger 2018 et al., to appear) 
• algorithm for mining high-utility itemsets in a transaction database containing negative unit profit values
  • the FHN algorithm (Fournier-Viger et al., 2014, powerpoint)
  • the HUINIV-Mine algorithm (Chu et al., 2009)
• algorithm for mining frequent high-utility itemsets in a transaction database
  • the FHMFreq algorithm, a variation of the FHM algorithm (Fournier-Viger et al., 2014)
• algorithm for mining on-shelf high-utility itemsets in a transaction database containing information about time periods of items
  • the FOSHU algorithm (Fournier-Viger et al., 2015, powerpoint)
  • the TS-HOUN algorithm (Lan et al., 2014)
• algorithm for incremental high-utility itemset mining in a transaction database
  
  • the EIHI algorithm (Fournier-Viger et al., 2015, powerpoint)
  • the HUI-LIST-INS algorithm (Lin et al., 2014)
• algorithm for mining concise representations of high-utility  itemsets in a transaction database
  • the HUG-Miner algorithm (Fournier-Viger et al., 2014, powerpoint) for mining high-utility generators
  • the GHUI-Miner algorithm (Fournier-Viger et al., 2014, powerpoint) for mining generators of high-utility itemsets
  • the MinFHM algorithm (Fournier-Viger et al., 2016, powerpoint, video ) for mining minimal high-utility itemsets
  • the EFIM-Closed algorithm (Fournier-Viger et al., 2016, powerpoint) for mining closed high-utility itemsets
  • the CHUI-Miner algorithm (Wu et al., 2015) for mining closed high-utility itemsets
  • the CHUD algorithm for mining closed high-utility itemsets (Tseng et al., 2011/2015)
  • the CHUI-Miner(Max) algorithm for mining maximal high utility itemsets (Wu et al., 2019). 
• algorithm for mining the skyline high-utility itemsets in a transaction database
  • the SkyMine algorithm (Goyal et al., 2015)
• algorithm for mining the top-k high-utility itemsets in a transaction database
  • the TKU algorithm (Tseng et al., 2015), obtained from UP-Miner under GPL license
  • the TKO-Basic algorithm (Tseng et al., 2015)
• algorithms for mining the top-k high utility itemsets from a data stream with a window
  • the FHMDS and FHMDS-Naive algorithms (Dawar et al. 2017) 
• algorithm for mining frequent skyline utility patterns in a transaction database
  • the SFUPMinerUemax algorithms (Lin et al, 2016) 
• algorithm for mining quantitative high utility itemsets in a transaction database:
  • the VHUQI algorithm (Wu et al., 2014)
• algorithm for mining high-utility sequential rules in a sequence database 
  • the HUSRM algorithm (Zida et al., 2015)
• algorithm for mining high-utility sequential patterns in a sequence database 
  • the USPAN algorithm (Yin et al. 2012)
• algorithm for mining high-utility probability sequential patterns in a sequence database 
  • the PHUSPM algorithm (Zhang et al. 2018) 
  • the UHUSPM algorithm (Zhang et al. 2018) 
• algorithm for mining high-utility itemsets in a transaction database using evolutionary algorithms
  • the HUIM-GA algorithm (Kannimuthu et al., 2014)
  • the HUIM-BPSO algorithm (Lin et al, 2016)
  • the HUIM-GA-tree algorithm (Lin et al, 2016)
  • the HUIM-BPSO-tree algorithm (Lin et al, 2016)
  • the HUIF-PSO algorithm (Song et al., 2018) 
  • the HUIF-GA algorithm (Song et al., 2018) 
  • the HUIF-BA algorithm (Song et al., 2018) 
• algorithm for mining high average-utility itemsets in a transaction database
  • the HAUI-Miner algorithm for mining high average-utility itemsets (Lin et al, 2016)
  • the EHAUPM algorithm for mining high average-utility itemsets (Lin et al, 2017) 
  • the HAUI-MMAU algorithm for mining high average-utility itemsets with multiple thresholds (Lin et al, 2016)
  • the MEMU algorithm for mining high average-utility itemsets with multiple thresholds (Lin et al, 2018)
• algorithms for mining high utility episodes in a sequence of complex events (a transaction database)
  • the TUP algorithm (Rathore et al., 2016) for mining frequent periodic patterns in a sequence of transactions (a transaction database))
  • the UP-SPAN algorithm (Wu et al., 2013 ) for mining periodic high-utility patterns (periodic patterns that yield a high profit) in a sequence of transactions (a transaction database) containing utility information 
• algorithms for mining periodic high-utility patterns (periodic patterns that yield a high profit) in a sequence of transactions (a transaction database) containing utility information
  • the PHM algorithm (Fournier-Viger et al, 2016b, powerpoint)
• algorithms for discovering irregular high utility itemsets (non periodic patterns) in a transaction database with utility information
  • the PHM_irregular algorithm, which is a simple variation of the PHM algorithm 
• algorithm for discovering local high utility itemsets in a database with utility information and timestamps
  • the LHUI-Miner algorithm (Fournier-Viger et al., 2019) 
• algorithm for discovering peak high utility itemsets in a database with utility information and timestamps
  • the PHUI-Miner algorithm (Fournier-Viger et al., 2019) 

Association Rule Mining

These algorithms discover interesting associations between symbols (values) in a transaction database (database records with binary attributes).

• an algorithm for mining all association rules in a transaction database (Agrawal & Srikant, 1994)
• an algorithm for mining all association rules with the lift measure in a transaction database (adapted from Agrawal & Srikant, 1994)
• an algorithm for mining the IGB informative and generic basis of association rules in a transaction database (Gasmi et al., 2005)
• an algorithm for mining perfectly sporadic association rules (Koh & Roundtree, 2005)
• an algorithm for mining closed association rules (Szathmary et al. 2006).
• an algorithm for mining minimal non redundant association rules (Kryszkiewicz, 1998)
• the Indirect algorithm for mining indirect association rules (Tan et al. 2000; Tan et 2006)
• the FHSAR algorithm for hiding sensitive association rules (Weng et al. 2008)
• the TopKRules algorithm for mining the top-k association rules (Fournier-Viger, 2012b, powerpoint)
• the TopKClassRules algorithm for mining the top-k class association rules (a variation of TopKRules. This latter is described in Fournier-Viger, 2012b, powerpoint)
• the TNR algorithm for mining top-k non-redundant association rules (Fournier-Viger 2012d, powerpoint)

Stream pattern mining

These algorithms discovers various kinds of patterns in a stream (an infinite sequence of database records (transactions))

• the estDec algorithm for mining recent frequent itemsets in a data stream (Chang & Lee, 2003)
• the estDec+ algorithm for mining recent frequent itemsets in a data stream (Shin et al., 2014)
• the CloStream algorithm for mining frequent closed itemsets in a data stream (Yen et al, 2009)
• algorithms for mining the top-k high utility itemsets from a data stream with a window
  • the FHMDS and FHMDS-Naive algorithms (Dawar et al. 2017) 

Clustering

These algorithms automatically find clusters in different kinds of data

• the original K-Means algorithm (MacQueen, 1967)
• the Bisecting K-Means algorithm (Steinbach et al, 2000)
• algorithms for density-based clustering
  • the DBScan algorithm (Ester et al., 1996)
  • the Optics algorithm to extract a cluster ordering of points, which can then be use to generate DBScan style clusters and more (Ankerst et al, 1999)
• a hierarchical clustering algorithm
• a tool called Cluster Viewer for visualizing clusters
• a tool called Instance Viewer for visualizing the input of clustering algorithms

Time series mining

These algorithms perform various tasks to analyze time series data

• an algorithm for converting a time series to a sequence of symbols using the SAX representation of time series. Note that if one converts a set of time series with SAX, he will obtain a sequence database, which allows to then apply traditional algorihtms for sequential rule mining and sequential pattern mining on time series (SAX, 2007).
• algorithms for calculating the prior moving average of a time series (to remove noise)
• algorithms for calculating the cumulative moving average f a time series (to remove noise)
• algorithms for calculating the central moving average of a time series (to remove noise)
• an algorithm for calculating the median smoothing of a time series (to remove noise)
• an algorithm for calculating the exponential smoothing of a time series (to remove noise) 
• an algorithm for calculating the min max normalization of a time series 
• an algorithm for calculating the autocorrelation function of a time series 
• an algorithm for calculating the standardization of a time series 
• an algorithm for calculating the first and second order differencing of a time series
• an algorithm for calculating the piecewise aggregate approximation of a time series (to reduce the number of data points of a time series)
• an algorithm for calculating the linear regression of a time series (using the least squares method) 
• an algorithm for splitting a time series into segments of a given length
• an algorithm for splitting a time series into a given number of segments
• algorithms to cluster time series (group time-series according to their similarities). This can be done by applying the clustering algorithms offered in SPMF (K-Means, Bisecting K-Means, DBScan, OPTICS, Hierarchical clustering) on time series.
• a tool called Time Series Viewer for visualizing time series